4 Two counterexamples to IEM-FP?

IEM is generally considered to be a property of judgments concerning the first-person perspective and respectively involving the first-person pronoun. A major problem in the current discussion about IEM is that no solid account of what judgments are is given. In contrast to philosophers that are concerned with beliefs, who usually give a brief declaration of what they take beliefs to be (e.g., relations, sentence operations etc.), philosophers involved in the IEM discussion seem to take judgments to be already widely understood. Since the initial paper written by Shoemaker (1968) focuses on the identification-freedom of judgments, we think that what philosophers usually talk about using the term “judgment” is inference or reasoning.[5]

So we take judgments to consist of propositional reasoning. Let us have a look at some examples:

Judgment A:

(1) John is a fish. (Fa)

(2) Fish can swim. (∀x)(Fx→Gx)

Judgment B:

(1) John is a fish. (Fa)

(2) John is Jim's best friend. (a=b)

Though usually the conclusion of these inferences is what is referred to using the term “judgment”, we do not think that philosophers generally tend to take judgments as being adequately analysed as propositional attitudes (as relations between persons and propositions like “Jim believes that it is raining”). We take the whole inference to be what is referred to with the term “judgments”, and the conclusion to be what is referred to using the term “belief”.[6] Judgments A and B are analogous in the following sense: they are both judgment involving two premises and their logically necessary conclusions. But they differ in a particular aspect that is of the highest importance concerning IEM: only judgment B involves an identification, whereas judgment A is identification-free. So the first thing we can say is that IEM following from identification-freedom is not an exclusive property of judgments involving the first-person pronoun—there are numerous judgments that do not contain any identification-components. This is our first reason for thinking that IEM may hold, but is not a remarkable or significant property exclusively reserved for judgments involving the first-person pronoun.

What Shoemaker wants to make clear is that there are certain judgments that cannot take the logical form of judgment B and that these judgments involve the first-person pronoun, in the sense of Wittgenstein’s “subject-use”. Let us take a look at what Shoemaker means by giving examples for the object-use and the subject-use:

Object-use:

(1) The person in the mirror is looking tired. (Fa)

(2) I am the person in the mirror. (a=b)

As you can see, there are judgments involving the first-person pronoun that also involve an identification-component—at least that is what Shoemaker (1968) thinks. But when he claims that there are judgments that are immune to error through misidentification, he does not claim that they are immune to any error, and nor does he claim that the identity relation holds with metaphysical necessity—he just claims that whenever one judges and this judgment involves certain kind of predicates (or properties) it is automatically identification-free. One of those predicates (or properties) is being in pain. Let's have a look at how the judgment would work with this special predicate that we may call P*.

Subject-use:

(1) There is something that is in pain. (∃x)(P*x)

(2) P* is always a property of the person recognizing it. (P*gen)

In fact this formal representation of such a judgment is even weaker than what Shoemaker may have had in mind, thus the strong reading of his idea of judgments that are IEM because of their identification freedom would be:

(1) There is something that is in pain. (∃x)(P*x)

This reading gets closer to Shoemaker's idea, because he would not agree that judgments explicitly involve a generalization such as (2). We undertook this brief exercise first of all to put pressure on the following point: although philosophers of different generations have been talking about IEM for decades, they usually fail to give an explicit account of what judgments are and how they work.[7] This exercise was meant to fill this theoretical gap for the purpose of the current discussion. So whenever someone utters “John is a fisherman”, we take this sentence to express a propositional attitude—a belief. But when we believe that he judges “John is a fisherman”, we also take that person to have made an inference, simply because that is what we want to say when we ascribe a judgment to him. Shoemaker's claim that judgments like “I am in pain” are immune to error through misidentification does not mean that there are hidden structures, neither of the sentence expressing the judgment nor of the propositional attitude expressed by the sentence “I am in pain”; it means that no identification-component was involved in the inference that has been made.

The second reason for undertaking this exercise is that we want to have a look at whether Liang's counterexamples (especially the somatoparaphrenia example) are real counterexamples. We do not think that the two examples Liang gives are in any way counterexamples to IEM—though they are philosophically very interesting, especially concerning theories of self-consciousness. Liang claims that the two counterexamples falsify the Immunity Principle, but we claim that they do not meet the conditions that have to be met to falsify this theory. So we must first see what Liang takes to falsify the IEM theory and then settle on a criterion for how the IEM theory could be falsified.

Liang thinks that the following would suffice for IEM to hold:

(1) for every phenomenal state there must be a subject who experiences it; (2) every phenomenal state is in principle available to first-personal access (Shoemaker 1996); (3) every phenomenal state is experienced by the one who has first-personal access to that state. The crucial point is that (1)–(3) do not imply that (4) every phenomenal state is, from the first-person point of view, represented as experienced by the one who has first-personal access to that state. (Liang this collection, p. 8)

Liang also considers his two counterexamples (the somatoparaphrenia patient and the body-swap illusion) to be counterexamples to (4), so the IEM-principle does not hold. In fact we agree with Liang that at least one of these examples is a counterexample to (4) but we do not agree that (4) is necessary for IEM to hold. So let us first have a look at how a falsification of the IEM-theory would have to look. The IEM-theory comes in the form of a material conditional: if a person judges “I am φ”, then she cannot be wrong because of a misidentification. The truth conditions for a material conditional are clear: the conditional is wrong if and only if the antecedent is true and the consequent is wrong. This brings us to the definition of a theoretical falsification of IEM:

Falsification of IEM =_Df : 1. The IEM-theory would be falsified if and only if a person judges “I am φ” and is wrong in her judgment because of a misidentification. Or, more precisely: 2. The IEM-theory would be falsified if and only if there is an example of a person that believes “I am φ” and comes to this belief through inference (the judgment) that involves an identification component and this identification is wrong.

Thus, speaking more formally, a judgment of the following two forms must be present (cf. Pryor 1999):

wh-judgment[8]

(1) (∃x)(Fx) (predication to a variable)

(2) I am x (identification, x=a)

de re judgment[9]

(1) A particular thing (de re) is F (Fa)

(2) I am that particular thing (a=b)

We pick these two different structures to emphasise that Shoemaker did not exclusively talk about de re “attitudes” but also about “existential quantification”, though he did not do so explicitly. Note that besides the presence of belief states such as (c) it is necessary for the falsification of IEM-theory that this conclusion is only wrong because (2) is wrong.

The crucial question concerning Liang's counterexamples is: do they meet this condition? Consider the first example. It is—as Liang sees it according to Pryor—an example of a de re misidentification. De re misidentifications occur, for example, when there are two objects equally eligible for exemplifying the property in question. To show that Liang's first example is a counterexample to IEM one would have to prove, first, that the structure of a de re judgment as stated above holds, and second that the judgment is only wrong because the second premise is not true. Recall the first experiment: a patient suffering from somatoparaphrenia, FB, is touched on her hand and asked whether she feels her hand being touched. She answers “No”. When she is asked whether she feels her niece's hand being touched, she gives a positive answer (FB believes that her hand is in fact her niece's hand, and has been placed on her body). But since she does not judge “I am being touched on my hand”, the necessary conditions for falsifying the IEM-theory are not met. The material conditional could only be proved wrong if the antecedent (a person judging that she has a certain property) is true, but in this case it is not true. The conditions would have been met if she had answered “I feel being touched on my hand”, even though she was not, and even though the only reason why she was wrong was because she misidentified her own sensations with someone else's. But she does not commit the error of judging “I am being touched” in the first place, so the IEM-theory is not falsified. It is crucial here to understand that falsification of IEM does not depend on what exactly she said, but whether she judged that she had a certain property. Unfortunately wrongly judging that one is not touched, though one is touched, does not get close to a falsification of IEM, by definition of the truth conditions of material conditionals.

Now let us have a look at Liang's second counterexample: the body-swap illusion. This, according to Liang and Pryor, is an example of a wh-misidentification that happens when someone simply knows a property to be there (e.g., a smell) and falsely ascribes this property to a particular object. In this setup, the participants judge that they are shaking hands with themselves. This example gets much closer to the claim of IEM, because they in fact judge, and judge falsely, that they experience something, and there is another person who really seems to have that experience. So it seems that one of the following inferences is made:

(1) A particular person is shaking hands with myself. (Fa)

(2) I am that person. (a=b)

(1) There is something that is shaking hands with myself. (∃x)(Fx)

(2) I am that something. (x=a)

If these judgments occurred it is obvious that they are false because the second premise is false—thus an error through misidentification was made. But did the participants really commit such an error? Recall that the IEM thesis would be falsified if a person believed a certain proposition but was mistaken because and only because she misidentified herself with someone else. Did the participants really believe that they were shaking hands with themselves? We assume that they most certainly did not. Of course they remarked that they were shaking hands with themselves, but we take them to speak merely metaphorically and not literally. If one wanted to be sure, the same experiment would have to be made, asking the participants whether they believed that they were shaking hand with themselves (not if it felt as if they were). Even if they believed that they were shaking hands with themselves, the judgment would probably not have the form stated above, because they did not have the experience of the other participant—only if they had an experience that that very (exactly the same) experience depended upon another person, and only if they accidently identified themselves with that person—only in that case would the IEM-theory be falsified. But the participants did not have the experience of the other person wearing the camera. They were having their very own experience—caused by the informational flow starting with the display (monitoring not the perspective of the person wearing the camera, but the camera's perspective) and their own lenses, their own retina, and so on. The experience they ascribed to themselves was not the experience of another person or agent, it was their own experience. They were only wrong in judging that they were shaking hands with themselves because they in fact did not shake hands with themselves. This is not a misidentification but simply a mispredication. This problem will be elaborated in section 5.

So why does Liang think that these examples are counterexamples to IEM? Because he takes (4) to be crucial for IEM to hold. The differences between what Shoemaker and Evans take to be the theory of IEM and what Liang takes it to be are the following:

Shoemaker/Evans: If a person believes that she has certain properties, she cannot be mistaken in having them by misidentifying herself (or her phenomenal states) with someone else or someone else's states. In this conditional, the antecedent implies a person to believe something about herself or, speaking in Liang's terms, a person to represent herself as having a so-and-so experience. But Liang's conditional looks quite different:

Liang: “(4) every phenomenal state is, from the first-person point of view, represented as experienced by the one who has first-personal access to that state.” (Liang this collection, p. 8)

So what used to be the antecedent in the original theory becomes the consequent in Liang's theory—thus Liang is right that (4) does not hold and that the somatoparaphrenia patient and her reports are counterexamples to (4), but he is not right in taking this fact to falsify the IEM-theory.